Atlas record: https://www.aitimeline.dev/en/events/evt-0104/
Milestone · 1965
The DENDRAL project begins
Stanford began building a system that identifies the structure of an organic molecule from its mass spectrum, using rules supplied by chemists.
Why it matters
The field recovered results by trading general search for a narrow task filled with expert knowledge.
DENDRAL demonstrated what Feigenbaum later stated outright: a system's power comes from domain knowledge, not from the elegance of its inference mechanism. Rules were extracted from chemists in long interviews, and that labour created the separate problem of knowledge acquisition. Meta-DENDRAL later tried to derive the rules automatically.
Event record
- Event date
- 1965 · Approximate date
- Timeline date
- Event date
- Verification
- Sources gathered automatically · September 27, 2026
- Lines
- ID
- evt-0104
The year 1965 follows the participants' retrospective; sources do not give a month. Lederberg's report to NASA, DENDRAL-64, is dated 15 December 1964: the notation algorithm the project grew from.
Related events
Earlier
Later
- Builds on MYCIN 1974links to this record
The system was built by the same people on the same domain-knowledge principle as DENDRAL.
- Extends AM 1976links to this record
The same Stanford approach, but the heuristics search for new concepts rather than a molecular structure.
- Builds on The Art of Artificial Intelligence 1977links to this record
The thesis generalises the experience of DENDRAL and MYCIN, which it presents as case studies.